Scholarlink

<p>Hyper-fast Lithium-Ion battery charging is limited by rapid local heating, electrolyte depletion, and nonuniform current distribution characteristics, which increase degradation and risk. By decoupling boundary dynamics, electrothermal and physics-informed models ignore fast transient thermal resistance, contact variability, and 3D electrolyte-phase transport. To fill this gap, we offer a Physics-Guided Neural Thermal Model with heat generation, electrochemical transport, and dynamic boundary conditions. Five steps form the framework’s analytical pipeline. Infrared thermography allows SPIN-FLOW to infer real-time convective and contact boundary fields. Physics-informed neural operator BENTO-PINN predicts 3D temperature and electrolyte concentration fields with unknown boundaries. To ensure conservation-law compatibility, the ECTT module validates and corrects these fields using weak-form PDE auditing. DRACO-UFC (Distributionally-Robust Adjoint-Free Control) optimizes charging methods by minimizing hotspot formation and temperature rise under stochastic boundary distributions. Finally, HiLo-RECAP provides adaptive correction and closed-loop experimental validation. Combining these methods results in RMSE &lt; 1.6&#xa0;K, hotspot reduction &gt; 40%, and consistent safety compliance across different boundary circumstances. The connected electro-thermal design allows real-time, uncertainty-aware ultra-fast charging control, enhancing battery digital twin fidelity and enabling safe, high-rate operation for next-generation energy systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Physics-guided neural thermal framework for real-time electro-thermal optimization in ultra-fast charging lithium-ion cells

  • 期刊论文

  • 发表日期：

    2026年4月23日

查看全文
  • Manisha Lande,
  • Devesh Shrivastava,
  • Pravin Khope,
  • Shashi Bahl,
  • Sameer S. Gajghate,
  • Sagar Shelare,
  • Muhamad M. Noor

摘要

Hyper-fast Lithium-Ion battery charging is limited by rapid local heating, electrolyte depletion, and nonuniform current distribution characteristics, which increase degradation and risk. By decoupling boundary dynamics, electrothermal and physics-informed models ignore fast transient thermal resistance, contact variability, and 3D electrolyte-phase transport. To fill this gap, we offer a Physics-Guided Neural Thermal Model with heat generation, electrochemical transport, and dynamic boundary conditions. Five steps form the framework’s analytical pipeline. Infrared thermography allows SPIN-FLOW to infer real-time convective and contact boundary fields. Physics-informed neural operator BENTO-PINN predicts 3D temperature and electrolyte concentration fields with unknown boundaries. To ensure conservation-law compatibility, the ECTT module validates and corrects these fields using weak-form PDE auditing. DRACO-UFC (Distributionally-Robust Adjoint-Free Control) optimizes charging methods by minimizing hotspot formation and temperature rise under stochastic boundary distributions. Finally, HiLo-RECAP provides adaptive correction and closed-loop experimental validation. Combining these methods results in RMSE < 1.6 K, hotspot reduction > 40%, and consistent safety compliance across different boundary circumstances. The connected electro-thermal design allows real-time, uncertainty-aware ultra-fast charging control, enhancing battery digital twin fidelity and enabling safe, high-rate operation for next-generation energy systems.